arXiv:2505.20349v2 Announce Type: replace-cross Abstract: Data-driven modeling of fluid dynamics has advanced rapidly with neural PDE solvers, yet a fair and strong benchmark remains fragmented due to the absence of unified PDE datasets and standardized evaluation protocols. Although architectural innovations are abundant, fair assessment is further impeded by the lack of clear disentanglement between spatial, temporal and loss modules. In this paper, we introduce FD-Bench, the first fair, modular, comprehensive and reproducible benchmark for data-driven fluid simulation. FD-Bench systematical
The rapid advancement of neural PDE solvers in data-driven fluid dynamics necessitates a standardized benchmark to consolidate progress and facilitate fair comparison among emerging architectural innovations.
A robust and fair benchmark like FD-Bench is crucial for accelerating research and development in AI-driven scientific simulation, which has widespread applications across various industries.
The introduction of FD-Bench provides a unified dataset and standardized evaluation protocols for data-driven fluid simulation, enabling clearer disentanglement and assessment of different AI model components.
- · AI researchers in scientific computing
- · Engineering design firms
- · Climate modeling institutions
- · Aerospace industry
- · Fragmented, ad-hoc fluid simulation methods
- · AI models lacking robust validation
- · Organizations relying on disparate datasets
Improved accuracy and efficiency of data-driven fluid simulations.
Faster development and deployment of AI models for complex physical systems.
Potential for AI to revolutionize design, optimization, and predictive modeling across diverse scientific and engineering disciplines.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG